From f126253040654d52db134063a69ebaf0c417d410 Mon Sep 17 00:00:00 2001 From: =?utf8?q?Marek=20Ol=C5=A1=C3=A1k?= Date: Tue, 31 Jan 2012 10:50:51 +0100 Subject: [PATCH] r600g: turn init_config into a command buffer for starting a CS This is the first pure command buffer. It contains CS initialization packets and emits invariant state (i.e. the registers which never or rarely change). The affected registers are removed from *_hw_context.c, so that both ways of emitting commands can co-exist. v2: emit context_control in cayman's start_cs too --- .../drivers/r600/evergreen_hw_context.c | 141 +----- src/gallium/drivers/r600/evergreen_state.c | 411 +++++++++--------- src/gallium/drivers/r600/evergreend.h | 1 - src/gallium/drivers/r600/r600_hw_context.c | 63 +-- .../drivers/r600/r600_hw_context_priv.h | 1 - src/gallium/drivers/r600/r600_pipe.c | 5 +- src/gallium/drivers/r600/r600_pipe.h | 67 ++- src/gallium/drivers/r600/r600_state.c | 142 +++--- src/gallium/drivers/r600/r600_state_common.c | 24 + src/gallium/drivers/r600/r600d.h | 3 - 10 files changed, 379 insertions(+), 479 deletions(-) diff --git a/src/gallium/drivers/r600/evergreen_hw_context.c b/src/gallium/drivers/r600/evergreen_hw_context.c index b24f168a841..00fdade3332 100644 --- a/src/gallium/drivers/r600/evergreen_hw_context.c +++ b/src/gallium/drivers/r600/evergreen_hw_context.c @@ -34,33 +34,11 @@ static const struct r600_reg evergreen_config_reg_list[] = { {R_008958_VGT_PRIMITIVE_TYPE, 0}, - {R_008A14_PA_CL_ENHANCE, REG_FLAG_FLUSH_CHANGE, 0}, - {R_008C00_SQ_CONFIG, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_008C04_SQ_GPR_RESOURCE_MGMT_1, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_008C08_SQ_GPR_RESOURCE_MGMT_2, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_008C0C_SQ_GPR_RESOURCE_MGMT_3, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_008C10_SQ_GLOBAL_GPR_RESOURCE_MGMT_1, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_008C14_SQ_GLOBAL_GPR_RESOURCE_MGMT_2, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_008C18_SQ_THREAD_RESOURCE_MGMT_1, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_008C1C_SQ_THREAD_RESOURCE_MGMT_2, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_008C20_SQ_STACK_RESOURCE_MGMT_1, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_008C24_SQ_STACK_RESOURCE_MGMT_2, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_008C28_SQ_STACK_RESOURCE_MGMT_3, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_008E2C_SQ_LDS_RESOURCE_MGMT, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_009100_SPI_CONFIG_CNTL, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_00913C_SPI_CONFIG_CNTL_1, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, }; static const struct r600_reg cayman_config_reg_list[] = { {R_008958_VGT_PRIMITIVE_TYPE, 0, 0}, - {R_008A14_PA_CL_ENHANCE, REG_FLAG_FLUSH_CHANGE, 0}, - {R_008C00_SQ_CONFIG, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_008C04_SQ_GPR_RESOURCE_MGMT_1, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_008C10_SQ_GLOBAL_GPR_RESOURCE_MGMT_1, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_008C14_SQ_GLOBAL_GPR_RESOURCE_MGMT_2, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, {R_009100_SPI_CONFIG_CNTL, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, {R_00913C_SPI_CONFIG_CNTL_1, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, }; @@ -125,38 +103,6 @@ static const struct r600_reg evergreen_context_reg_list[] = { {R_0282D0_PA_SC_VPORT_ZMIN_0, 0, 0}, {R_0282D4_PA_SC_VPORT_ZMAX_0, 0, 0}, {R_028350_SX_MISC, 0, 0}, - {R_028380_SQ_VTX_SEMANTIC_0, 0, 0}, - {R_028384_SQ_VTX_SEMANTIC_1, 0, 0}, - {R_028388_SQ_VTX_SEMANTIC_2, 0, 0}, - {R_02838C_SQ_VTX_SEMANTIC_3, 0, 0}, - {R_028390_SQ_VTX_SEMANTIC_4, 0, 0}, - {R_028394_SQ_VTX_SEMANTIC_5, 0, 0}, - {R_028398_SQ_VTX_SEMANTIC_6, 0, 0}, - {R_02839C_SQ_VTX_SEMANTIC_7, 0, 0}, - {R_0283A0_SQ_VTX_SEMANTIC_8, 0, 0}, - {R_0283A4_SQ_VTX_SEMANTIC_9, 0, 0}, - {R_0283A8_SQ_VTX_SEMANTIC_10, 0, 0}, - {R_0283AC_SQ_VTX_SEMANTIC_11, 0, 0}, - {R_0283B0_SQ_VTX_SEMANTIC_12, 0, 0}, - {R_0283B4_SQ_VTX_SEMANTIC_13, 0, 0}, - {R_0283B8_SQ_VTX_SEMANTIC_14, 0, 0}, - {R_0283BC_SQ_VTX_SEMANTIC_15, 0, 0}, - {R_0283C0_SQ_VTX_SEMANTIC_16, 0, 0}, - {R_0283C4_SQ_VTX_SEMANTIC_17, 0, 0}, - {R_0283C8_SQ_VTX_SEMANTIC_18, 0, 0}, - {R_0283CC_SQ_VTX_SEMANTIC_19, 0, 0}, - {R_0283D0_SQ_VTX_SEMANTIC_20, 0, 0}, - {R_0283D4_SQ_VTX_SEMANTIC_21, 0, 0}, - {R_0283D8_SQ_VTX_SEMANTIC_22, 0, 0}, - {R_0283DC_SQ_VTX_SEMANTIC_23, 0, 0}, - {R_0283E0_SQ_VTX_SEMANTIC_24, 0, 0}, - {R_0283E4_SQ_VTX_SEMANTIC_25, 0, 0}, - {R_0283E8_SQ_VTX_SEMANTIC_26, 0, 0}, - {R_0283EC_SQ_VTX_SEMANTIC_27, 0, 0}, - {R_0283F0_SQ_VTX_SEMANTIC_28, 0, 0}, - {R_0283F4_SQ_VTX_SEMANTIC_29, 0, 0}, - {R_0283F8_SQ_VTX_SEMANTIC_30, 0, 0}, - {R_0283FC_SQ_VTX_SEMANTIC_31, 0, 0}, {GROUP_FORCE_NEW_BLOCK, 0, 0}, {R_028400_VGT_MAX_VTX_INDX, 0, 0}, {R_028404_VGT_MIN_VTX_INDX, 0, 0}, @@ -273,7 +219,6 @@ static const struct r600_reg evergreen_context_reg_list[] = { {R_028818_PA_CL_VTE_CNTL, 0, 0}, {R_02881C_PA_CL_VS_OUT_CNTL, 0, 0}, {R_028820_PA_CL_NANINF_CNTL, 0, 0}, - {R_028838_SQ_DYN_GPR_RESOURCE_LIMIT_1, 0, 0}, {R_028840_SQ_PGM_START_PS, REG_FLAG_NEED_BO, 0}, {R_028844_SQ_PGM_RESOURCES_PS, 0, 0}, {R_028848_SQ_PGM_RESOURCES_2_PS, 0, 0}, @@ -284,16 +229,6 @@ static const struct r600_reg evergreen_context_reg_list[] = { {R_0288A4_SQ_PGM_START_FS, REG_FLAG_NEED_BO, 0}, {R_0288A8_SQ_PGM_RESOURCES_FS, 0, 0}, {R_0288EC_SQ_LDS_ALLOC_PS, 0, 0}, - {R_028900_SQ_ESGS_RING_ITEMSIZE, 0, 0}, - {R_028904_SQ_GSVS_RING_ITEMSIZE, 0, 0}, - {R_028908_SQ_ESTMP_RING_ITEMSIZE, 0, 0}, - {R_02890C_SQ_GSTMP_RING_ITEMSIZE, 0, 0}, - {R_028910_SQ_VSTMP_RING_ITEMSIZE, 0, 0}, - {R_028914_SQ_PSTMP_RING_ITEMSIZE, 0, 0}, - {R_02891C_SQ_GS_VERT_ITEMSIZE, 0, 0}, - {R_028920_SQ_GS_VERT_ITEMSIZE_1, 0, 0}, - {R_028924_SQ_GS_VERT_ITEMSIZE_2, 0, 0}, - {R_028928_SQ_GS_VERT_ITEMSIZE_3, 0, 0}, {R_028940_ALU_CONST_CACHE_PS_0, REG_FLAG_NEED_BO, 0}, {R_028944_ALU_CONST_CACHE_PS_1, REG_FLAG_NEED_BO, 0}, {R_028980_ALU_CONST_CACHE_VS_0, REG_FLAG_NEED_BO, 0}, @@ -302,23 +237,7 @@ static const struct r600_reg evergreen_context_reg_list[] = { {R_028A04_PA_SU_POINT_MINMAX, 0, 0}, {R_028A08_PA_SU_LINE_CNTL, 0, 0}, {R_028A0C_PA_SC_LINE_STIPPLE, 0, 0}, - {R_028A10_VGT_OUTPUT_PATH_CNTL, 0, 0}, - {R_028A14_VGT_HOS_CNTL, 0, 0}, - {R_028A18_VGT_HOS_MAX_TESS_LEVEL, 0, 0}, - {R_028A1C_VGT_HOS_MIN_TESS_LEVEL, 0, 0}, - {R_028A20_VGT_HOS_REUSE_DEPTH, 0, 0}, - {R_028A24_VGT_GROUP_PRIM_TYPE, 0, 0}, - {R_028A28_VGT_GROUP_FIRST_DECR, 0, 0}, - {R_028A2C_VGT_GROUP_DECR, 0, 0}, - {R_028A30_VGT_GROUP_VECT_0_CNTL, 0, 0}, - {R_028A34_VGT_GROUP_VECT_1_CNTL, 0, 0}, - {R_028A38_VGT_GROUP_VECT_0_FMT_CNTL, 0, 0}, - {R_028A3C_VGT_GROUP_VECT_1_FMT_CNTL, 0, 0}, - {R_028A40_VGT_GS_MODE, 0, 0}, {R_028A48_PA_SC_MODE_CNTL_0, 0, 0}, - {R_028A4C_PA_SC_MODE_CNTL_1, 0, 0}, - {R_028AB4_VGT_REUSE_OFF, 0, 0}, - {R_028AB8_VGT_VTX_CNT_EN, 0, 0}, {R_028ABC_DB_HTILE_SURFACE, 0, 0}, {R_028AC0_DB_SRESULTS_COMPARE_STATE0, 0, 0}, {R_028AC4_DB_SRESULTS_COMPARE_STATE1, 0, 0}, @@ -331,8 +250,6 @@ static const struct r600_reg evergreen_context_reg_list[] = { {R_028B84_PA_SU_POLY_OFFSET_FRONT_OFFSET, 0, 0}, {R_028B88_PA_SU_POLY_OFFSET_BACK_SCALE, 0, 0}, {R_028B8C_PA_SU_POLY_OFFSET_BACK_OFFSET, 0, 0}, - {R_028B94_VGT_STRMOUT_CONFIG, 0, 0}, - {R_028B98_VGT_STRMOUT_BUFFER_CONFIG, 0, 0}, {R_028C00_PA_SC_LINE_CNTL, 0, 0}, {R_028C04_PA_SC_AA_CONFIG, 0, 0}, {R_028C08_PA_SU_VTX_CNTL, 0, 0}, @@ -495,38 +412,6 @@ static const struct r600_reg cayman_context_reg_list[] = { {R_0282D0_PA_SC_VPORT_ZMIN_0, 0, 0}, {R_0282D4_PA_SC_VPORT_ZMAX_0, 0, 0}, {R_028350_SX_MISC, 0, 0}, - {R_028380_SQ_VTX_SEMANTIC_0, 0, 0}, - {R_028384_SQ_VTX_SEMANTIC_1, 0, 0}, - {R_028388_SQ_VTX_SEMANTIC_2, 0, 0}, - {R_02838C_SQ_VTX_SEMANTIC_3, 0, 0}, - {R_028390_SQ_VTX_SEMANTIC_4, 0, 0}, - {R_028394_SQ_VTX_SEMANTIC_5, 0, 0}, - {R_028398_SQ_VTX_SEMANTIC_6, 0, 0}, - {R_02839C_SQ_VTX_SEMANTIC_7, 0, 0}, - {R_0283A0_SQ_VTX_SEMANTIC_8, 0, 0}, - {R_0283A4_SQ_VTX_SEMANTIC_9, 0, 0}, - {R_0283A8_SQ_VTX_SEMANTIC_10, 0, 0}, - {R_0283AC_SQ_VTX_SEMANTIC_11, 0, 0}, - {R_0283B0_SQ_VTX_SEMANTIC_12, 0, 0}, - {R_0283B4_SQ_VTX_SEMANTIC_13, 0, 0}, - {R_0283B8_SQ_VTX_SEMANTIC_14, 0, 0}, - {R_0283BC_SQ_VTX_SEMANTIC_15, 0, 0}, - {R_0283C0_SQ_VTX_SEMANTIC_16, 0, 0}, - {R_0283C4_SQ_VTX_SEMANTIC_17, 0, 0}, - {R_0283C8_SQ_VTX_SEMANTIC_18, 0, 0}, - {R_0283CC_SQ_VTX_SEMANTIC_19, 0, 0}, - {R_0283D0_SQ_VTX_SEMANTIC_20, 0, 0}, - {R_0283D4_SQ_VTX_SEMANTIC_21, 0, 0}, - {R_0283D8_SQ_VTX_SEMANTIC_22, 0, 0}, - {R_0283DC_SQ_VTX_SEMANTIC_23, 0, 0}, - {R_0283E0_SQ_VTX_SEMANTIC_24, 0, 0}, - {R_0283E4_SQ_VTX_SEMANTIC_25, 0, 0}, - {R_0283E8_SQ_VTX_SEMANTIC_26, 0, 0}, - {R_0283EC_SQ_VTX_SEMANTIC_27, 0, 0}, - {R_0283F0_SQ_VTX_SEMANTIC_28, 0, 0}, - {R_0283F4_SQ_VTX_SEMANTIC_29, 0, 0}, - {R_0283F8_SQ_VTX_SEMANTIC_30, 0, 0}, - {R_0283FC_SQ_VTX_SEMANTIC_31, 0, 0}, {GROUP_FORCE_NEW_BLOCK, 0, 0}, {R_028400_VGT_MAX_VTX_INDX, 0, 0}, {R_028404_VGT_MIN_VTX_INDX, 0, 0}, @@ -633,7 +518,6 @@ static const struct r600_reg cayman_context_reg_list[] = { {R_028798_CB_BLEND6_CONTROL, 0, 0}, {R_02879C_CB_BLEND7_CONTROL, 0, 0}, {R_028800_DB_DEPTH_CONTROL, 0, 0}, - {CM_R_028804_DB_EQAA, 0, 0}, {R_028808_CB_COLOR_CONTROL, 0, 0}, {R_02880C_DB_SHADER_CONTROL, 0, 0}, {R_028810_PA_CL_CLIP_CNTL, 0, 0}, @@ -651,8 +535,6 @@ static const struct r600_reg cayman_context_reg_list[] = { {R_028864_SQ_PGM_RESOURCES_2_VS, 0, 0}, {R_0288A4_SQ_PGM_START_FS, REG_FLAG_NEED_BO, 0}, {R_0288A8_SQ_PGM_RESOURCES_FS, 0, 0}, - {CM_R_0288E8_SQ_LDS_ALLOC, 0, 0}, - {R_0288EC_SQ_LDS_ALLOC_PS, 0, 0}, {R_028900_SQ_ESGS_RING_ITEMSIZE, 0, 0}, {R_028904_SQ_GSVS_RING_ITEMSIZE, 0, 0}, {R_028908_SQ_ESTMP_RING_ITEMSIZE, 0, 0}, @@ -671,24 +553,7 @@ static const struct r600_reg cayman_context_reg_list[] = { {R_028A04_PA_SU_POINT_MINMAX, 0, 0}, {R_028A08_PA_SU_LINE_CNTL, 0, 0}, {R_028A0C_PA_SC_LINE_STIPPLE, 0, 0}, - {R_028A10_VGT_OUTPUT_PATH_CNTL, 0, 0}, - {R_028A14_VGT_HOS_CNTL, 0, 0}, - {R_028A18_VGT_HOS_MAX_TESS_LEVEL, 0, 0}, - {R_028A1C_VGT_HOS_MIN_TESS_LEVEL, 0, 0}, - {R_028A20_VGT_HOS_REUSE_DEPTH, 0, 0}, - {R_028A24_VGT_GROUP_PRIM_TYPE, 0, 0}, - {R_028A28_VGT_GROUP_FIRST_DECR, 0, 0}, - {R_028A2C_VGT_GROUP_DECR, 0, 0}, - {R_028A30_VGT_GROUP_VECT_0_CNTL, 0, 0}, - {R_028A34_VGT_GROUP_VECT_1_CNTL, 0, 0}, - {R_028A38_VGT_GROUP_VECT_0_FMT_CNTL, 0, 0}, - {R_028A3C_VGT_GROUP_VECT_1_FMT_CNTL, 0, 0}, - {R_028A40_VGT_GS_MODE, 0, 0}, {R_028A48_PA_SC_MODE_CNTL_0, 0, 0}, - {R_028A4C_PA_SC_MODE_CNTL_1, 0, 0}, - {CM_R_028AA8_IA_MULTI_VGT_PARAM, 0, 0}, - {R_028AB4_VGT_REUSE_OFF, 0, 0}, - {R_028AB8_VGT_VTX_CNT_EN, 0, 0}, {R_028ABC_DB_HTILE_SURFACE, 0, 0}, {R_028AC0_DB_SRESULTS_COMPARE_STATE0, 0, 0}, {R_028AC4_DB_SRESULTS_COMPARE_STATE1, 0, 0}, @@ -701,10 +566,6 @@ static const struct r600_reg cayman_context_reg_list[] = { {R_028B84_PA_SU_POLY_OFFSET_FRONT_OFFSET, 0, 0}, {R_028B88_PA_SU_POLY_OFFSET_BACK_SCALE, 0, 0}, {R_028B8C_PA_SU_POLY_OFFSET_BACK_OFFSET, 0, 0}, - {R_028B94_VGT_STRMOUT_CONFIG, 0, 0}, - {R_028B98_VGT_STRMOUT_BUFFER_CONFIG, 0, 0}, - {CM_R_028BD4_PA_SC_CENTROID_PRIORITY_0, 0, 0}, - {CM_R_028BD8_PA_SC_CENTROID_PRIORITY_1, 0, 0}, {CM_R_028BDC_PA_SC_LINE_CNTL, 0, 0}, {CM_R_028BE0_PA_SC_AA_CONFIG, 0, 0}, {CM_R_028BE4_PA_SU_VTX_CNTL, 0, 0}, @@ -995,8 +856,8 @@ int evergreen_context_init(struct r600_context *ctx) goto out_err; ctx->cs = ctx->ws->cs_create(ctx->ws); + r600_emit_atom(ctx, &ctx->atom_start_cs.atom); - r600_init_cs(ctx); ctx->max_db = 8; return 0; out_err: diff --git a/src/gallium/drivers/r600/evergreen_state.c b/src/gallium/drivers/r600/evergreen_state.c index be193a27789..2155fa0eac9 100644 --- a/src/gallium/drivers/r600/evergreen_state.c +++ b/src/gallium/drivers/r600/evergreen_state.c @@ -1803,91 +1803,105 @@ void evergreen_init_state_functions(struct r600_context *rctx) rctx->context.set_stream_output_targets = r600_set_so_targets; } -static void cayman_init_config(struct r600_context *rctx) +static void cayman_init_atom_start_cs(struct r600_context *rctx) { - struct r600_pipe_state *rstate = &rctx->config; - unsigned tmp; + struct r600_command_buffer *cb = &rctx->atom_start_cs; - tmp = 0x00000000; - tmp |= S_008C00_EXPORT_SRC_C(1); - r600_pipe_state_add_reg(rstate, R_008C00_SQ_CONFIG, tmp, NULL, 0); + r600_init_command_buffer(cb, 256, EMIT_EARLY); + + /* This must be first. */ + r600_store_value(cb, PKT3(PKT3_CONTEXT_CONTROL, 1, 0)); + r600_store_value(cb, 0x80000000); + r600_store_value(cb, 0x80000000); + r600_store_config_reg_seq(cb, R_008C00_SQ_CONFIG, 2); + r600_store_value(cb, S_008C00_EXPORT_SRC_C(1)); /* R_008C00_SQ_CONFIG */ /* always set the temp clauses */ - r600_pipe_state_add_reg(rstate, R_008C04_SQ_GPR_RESOURCE_MGMT_1, S_008C04_NUM_CLAUSE_TEMP_GPRS(4), NULL, 0); - r600_pipe_state_add_reg(rstate, R_008C10_SQ_GLOBAL_GPR_RESOURCE_MGMT_1, 0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_008C14_SQ_GLOBAL_GPR_RESOURCE_MGMT_2, 0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, (1 << 8), NULL, 0); - - r600_pipe_state_add_reg(rstate, R_028A4C_PA_SC_MODE_CNTL_1, 0x0, NULL, 0); - - r600_pipe_state_add_reg(rstate, R_028A10_VGT_OUTPUT_PATH_CNTL, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A14_VGT_HOS_CNTL, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A18_VGT_HOS_MAX_TESS_LEVEL, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A1C_VGT_HOS_MIN_TESS_LEVEL, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A20_VGT_HOS_REUSE_DEPTH, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A24_VGT_GROUP_PRIM_TYPE, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A28_VGT_GROUP_FIRST_DECR, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A2C_VGT_GROUP_DECR, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A30_VGT_GROUP_VECT_0_CNTL, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A34_VGT_GROUP_VECT_1_CNTL, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A38_VGT_GROUP_VECT_0_FMT_CNTL, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A3C_VGT_GROUP_VECT_1_FMT_CNTL, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A40_VGT_GS_MODE, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028B94_VGT_STRMOUT_CONFIG, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028B98_VGT_STRMOUT_BUFFER_CONFIG, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, CM_R_028AA8_IA_MULTI_VGT_PARAM, S_028AA8_SWITCH_ON_EOP(1) | S_028AA8_PARTIAL_VS_WAVE_ON(1) | S_028AA8_PRIMGROUP_SIZE(63), NULL, 0); - r600_pipe_state_add_reg(rstate, R_028AB4_VGT_REUSE_OFF, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028AB8_VGT_VTX_CNT_EN, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_008A14_PA_CL_ENHANCE, (3 << 1) | 1, NULL, 0); - - r600_pipe_state_add_reg(rstate, R_028380_SQ_VTX_SEMANTIC_0, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028384_SQ_VTX_SEMANTIC_1, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028388_SQ_VTX_SEMANTIC_2, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_02838C_SQ_VTX_SEMANTIC_3, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028390_SQ_VTX_SEMANTIC_4, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028394_SQ_VTX_SEMANTIC_5, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028398_SQ_VTX_SEMANTIC_6, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_02839C_SQ_VTX_SEMANTIC_7, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283A0_SQ_VTX_SEMANTIC_8, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283A4_SQ_VTX_SEMANTIC_9, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283A8_SQ_VTX_SEMANTIC_10, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283AC_SQ_VTX_SEMANTIC_11, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283B0_SQ_VTX_SEMANTIC_12, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283B4_SQ_VTX_SEMANTIC_13, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283B8_SQ_VTX_SEMANTIC_14, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283BC_SQ_VTX_SEMANTIC_15, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283C0_SQ_VTX_SEMANTIC_16, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283C4_SQ_VTX_SEMANTIC_17, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283C8_SQ_VTX_SEMANTIC_18, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283CC_SQ_VTX_SEMANTIC_19, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283D0_SQ_VTX_SEMANTIC_20, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283D4_SQ_VTX_SEMANTIC_21, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283D8_SQ_VTX_SEMANTIC_22, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283DC_SQ_VTX_SEMANTIC_23, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283E0_SQ_VTX_SEMANTIC_24, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283E4_SQ_VTX_SEMANTIC_25, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283E8_SQ_VTX_SEMANTIC_26, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283EC_SQ_VTX_SEMANTIC_27, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283F0_SQ_VTX_SEMANTIC_28, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283F4_SQ_VTX_SEMANTIC_29, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283F8_SQ_VTX_SEMANTIC_30, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283FC_SQ_VTX_SEMANTIC_31, 0x0, NULL, 0); - - r600_pipe_state_add_reg(rstate, R_028810_PA_CL_CLIP_CNTL, 0x0, NULL, 0); - - r600_pipe_state_add_reg(rstate, CM_R_028BD4_PA_SC_CENTROID_PRIORITY_0, 0x76543210, NULL, 0); - r600_pipe_state_add_reg(rstate, CM_R_028BD8_PA_SC_CENTROID_PRIORITY_1, 0xfedcba98, NULL, 0); - - r600_pipe_state_add_reg(rstate, CM_R_0288E8_SQ_LDS_ALLOC, 0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0288EC_SQ_LDS_ALLOC_PS, 0, NULL, 0); - - r600_pipe_state_add_reg(rstate, CM_R_028804_DB_EQAA, 0x110000, NULL, 0); - r600_context_pipe_state_set(rctx, rstate); + r600_store_value(cb, S_008C04_NUM_CLAUSE_TEMP_GPRS(4)); /* R_008C04_SQ_GPR_RESOURCE_MGMT_1 */ + + r600_store_config_reg_seq(cb, R_008C10_SQ_GLOBAL_GPR_RESOURCE_MGMT_1, 2); + r600_store_value(cb, 0); /* R_008C10_SQ_GLOBAL_GPR_RESOURCE_MGMT_1 */ + r600_store_value(cb, 0); /* R_008C14_SQ_GLOBAL_GPR_RESOURCE_MGMT_2 */ + + r600_store_config_reg(cb, R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, (1 << 8)); + + r600_store_context_reg(cb, R_028A4C_PA_SC_MODE_CNTL_1, 0); + + r600_store_context_reg_seq(cb, R_028A10_VGT_OUTPUT_PATH_CNTL, 13); + r600_store_value(cb, 0); /* R_028A10_VGT_OUTPUT_PATH_CNTL */ + r600_store_value(cb, 0); /* R_028A14_VGT_HOS_CNTL */ + r600_store_value(cb, 0); /* R_028A18_VGT_HOS_MAX_TESS_LEVEL */ + r600_store_value(cb, 0); /* R_028A1C_VGT_HOS_MIN_TESS_LEVEL */ + r600_store_value(cb, 0); /* R_028A20_VGT_HOS_REUSE_DEPTH */ + r600_store_value(cb, 0); /* R_028A24_VGT_GROUP_PRIM_TYPE */ + r600_store_value(cb, 0); /* R_028A28_VGT_GROUP_FIRST_DECR */ + r600_store_value(cb, 0); /* R_028A2C_VGT_GROUP_DECR */ + r600_store_value(cb, 0); /* R_028A30_VGT_GROUP_VECT_0_CNTL */ + r600_store_value(cb, 0); /* R_028A34_VGT_GROUP_VECT_1_CNTL */ + r600_store_value(cb, 0); /* R_028A38_VGT_GROUP_VECT_0_FMT_CNTL */ + r600_store_value(cb, 0); /* R_028A3C_VGT_GROUP_VECT_1_FMT_CNTL */ + r600_store_value(cb, 0); /* R_028A40_VGT_GS_MODE */ + + r600_store_context_reg_seq(cb, R_028B94_VGT_STRMOUT_CONFIG, 2); + r600_store_value(cb, 0); /* R_028B94_VGT_STRMOUT_CONFIG */ + r600_store_value(cb, 0); /* R_028B98_VGT_STRMOUT_BUFFER_CONFIG */ + + r600_store_context_reg_seq(cb, R_028AB4_VGT_REUSE_OFF, 2); + r600_store_value(cb, 0); /* R_028AB4_VGT_REUSE_OFF */ + r600_store_value(cb, 0); /* R_028AB8_VGT_VTX_CNT_EN */ + + r600_store_config_reg(cb, R_008A14_PA_CL_ENHANCE, (3 << 1) | 1); + + r600_store_context_reg(cb, CM_R_028AA8_IA_MULTI_VGT_PARAM, S_028AA8_SWITCH_ON_EOP(1) | S_028AA8_PARTIAL_VS_WAVE_ON(1) | S_028AA8_PRIMGROUP_SIZE(63)); + + r600_store_context_reg_seq(cb, CM_R_028BD4_PA_SC_CENTROID_PRIORITY_0, 2); + r600_store_value(cb, 0x76543210); /* CM_R_028BD4_PA_SC_CENTROID_PRIORITY_0 */ + r600_store_value(cb, 0xfedcba98); /* CM_R_028BD8_PA_SC_CENTROID_PRIORITY_1 */ + + r600_store_context_reg_seq(cb, CM_R_0288E8_SQ_LDS_ALLOC, 2); + r600_store_value(cb, 0); /* CM_R_0288E8_SQ_LDS_ALLOC */ + r600_store_value(cb, 0); /* R_0288EC_SQ_LDS_ALLOC_PS */ + + r600_store_context_reg(cb, CM_R_028804_DB_EQAA, 0x110000); + + r600_store_context_reg_seq(cb, R_028380_SQ_VTX_SEMANTIC_0, 32); + r600_store_value(cb, 0); /* R_028380_SQ_VTX_SEMANTIC_0 */ + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); /* R_0283FC_SQ_VTX_SEMANTIC_31 */ } -void evergreen_init_config(struct r600_context *rctx) +void evergreen_init_atom_start_cs(struct r600_context *rctx) { - struct r600_pipe_state *rstate = &rctx->config; + struct r600_command_buffer *cb = &rctx->atom_start_cs; int ps_prio; int vs_prio; int gs_prio; @@ -1915,13 +1929,19 @@ void evergreen_init_config(struct r600_context *rctx) enum radeon_family family; unsigned tmp; - family = rctx->family; - if (rctx->chip_class == CAYMAN) { - cayman_init_config(rctx); + cayman_init_atom_start_cs(rctx); return; } - + + r600_init_command_buffer(cb, 256, EMIT_EARLY); + + /* This must be first. */ + r600_store_value(cb, PKT3(PKT3_CONTEXT_CONTROL, 1, 0)); + r600_store_value(cb, 0x80000000); + r600_store_value(cb, 0x80000000); + + family = rctx->family; ps_prio = 0; vs_prio = 1; gs_prio = 2; @@ -2145,7 +2165,7 @@ void evergreen_init_config(struct r600_context *rctx) break; } - tmp = 0x00000000; + tmp = 0; switch (family) { case CHIP_CEDAR: case CHIP_PALM: @@ -2165,150 +2185,145 @@ void evergreen_init_config(struct r600_context *rctx) tmp |= S_008C00_VS_PRIO(vs_prio); tmp |= S_008C00_GS_PRIO(gs_prio); tmp |= S_008C00_ES_PRIO(es_prio); - r600_pipe_state_add_reg(rstate, R_008C00_SQ_CONFIG, tmp, NULL, 0); /* enable dynamic GPR resource management */ if (rctx->screen->info.drm_minor >= 7) { + r600_store_config_reg_seq(cb, R_008C00_SQ_CONFIG, 2); + r600_store_value(cb, tmp); /* R_008C00_SQ_CONFIG */ /* always set temp clauses */ - r600_pipe_state_add_reg(rstate, R_008C04_SQ_GPR_RESOURCE_MGMT_1, - S_008C04_NUM_CLAUSE_TEMP_GPRS(num_temp_gprs), NULL, 0); - r600_pipe_state_add_reg(rstate, R_008C10_SQ_GLOBAL_GPR_RESOURCE_MGMT_1, 0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_008C14_SQ_GLOBAL_GPR_RESOURCE_MGMT_2, 0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, (1 << 8), NULL, 0); - r600_pipe_state_add_reg(rstate, R_028838_SQ_DYN_GPR_RESOURCE_LIMIT_1, + r600_store_value(cb, S_008C04_NUM_CLAUSE_TEMP_GPRS(num_temp_gprs)); /* R_008C04_SQ_GPR_RESOURCE_MGMT_1 */ + r600_store_config_reg_seq(cb, R_008C10_SQ_GLOBAL_GPR_RESOURCE_MGMT_1, 2); + r600_store_value(cb, 0); /* R_008C10_SQ_GLOBAL_GPR_RESOURCE_MGMT_1 */ + r600_store_value(cb, 0); /* R_008C14_SQ_GLOBAL_GPR_RESOURCE_MGMT_2 */ + r600_store_config_reg(cb, R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, (1 << 8)); + r600_store_context_reg(cb, R_028838_SQ_DYN_GPR_RESOURCE_LIMIT_1, S_028838_PS_GPRS(0x1e) | S_028838_VS_GPRS(0x1e) | S_028838_GS_GPRS(0x1e) | S_028838_ES_GPRS(0x1e) | S_028838_HS_GPRS(0x1e) | - S_028838_LS_GPRS(0x1e), NULL, 0); /* workaround for hw issues with dyn gpr - must set all limits to 240 instead of 0, 0x1e == 240 / 8*/ + S_028838_LS_GPRS(0x1e)); /* workaround for hw issues with dyn gpr - must set all limits to 240 instead of 0, 0x1e == 240 / 8*/ } else { - tmp = 0; - tmp |= S_008C04_NUM_PS_GPRS(num_ps_gprs); + r600_store_config_reg_seq(cb, R_008C00_SQ_CONFIG, 4); + r600_store_value(cb, tmp); /* R_008C00_SQ_CONFIG */ + + tmp = S_008C04_NUM_PS_GPRS(num_ps_gprs); tmp |= S_008C04_NUM_VS_GPRS(num_vs_gprs); tmp |= S_008C04_NUM_CLAUSE_TEMP_GPRS(num_temp_gprs); - r600_pipe_state_add_reg(rstate, R_008C04_SQ_GPR_RESOURCE_MGMT_1, tmp, NULL, 0); + r600_store_value(cb, tmp); /* R_008C04_SQ_GPR_RESOURCE_MGMT_1 */ - tmp = 0; - tmp |= S_008C08_NUM_GS_GPRS(num_gs_gprs); + tmp = S_008C08_NUM_GS_GPRS(num_gs_gprs); tmp |= S_008C08_NUM_ES_GPRS(num_es_gprs); - r600_pipe_state_add_reg(rstate, R_008C08_SQ_GPR_RESOURCE_MGMT_2, tmp, NULL, 0); + r600_store_value(cb, tmp); /* R_008C08_SQ_GPR_RESOURCE_MGMT_2 */ - tmp = 0; - tmp |= S_008C0C_NUM_HS_GPRS(num_hs_gprs); + tmp = S_008C0C_NUM_HS_GPRS(num_hs_gprs); tmp |= S_008C0C_NUM_HS_GPRS(num_ls_gprs); - r600_pipe_state_add_reg(rstate, R_008C0C_SQ_GPR_RESOURCE_MGMT_3, tmp, NULL, 0); + r600_store_value(cb, tmp); /* R_008C0C_SQ_GPR_RESOURCE_MGMT_3 */ } - tmp = 0; - tmp |= S_008C18_NUM_PS_THREADS(num_ps_threads); + tmp = S_008C18_NUM_PS_THREADS(num_ps_threads); tmp |= S_008C18_NUM_VS_THREADS(num_vs_threads); tmp |= S_008C18_NUM_GS_THREADS(num_gs_threads); tmp |= S_008C18_NUM_ES_THREADS(num_es_threads); - r600_pipe_state_add_reg(rstate, R_008C18_SQ_THREAD_RESOURCE_MGMT_1, tmp, NULL, 0); + r600_store_config_reg_seq(cb, R_008C18_SQ_THREAD_RESOURCE_MGMT_1, 5); + r600_store_value(cb, tmp); /* R_008C18_SQ_THREAD_RESOURCE_MGMT_1 */ - tmp = 0; - tmp |= S_008C1C_NUM_HS_THREADS(num_hs_threads); + tmp = S_008C1C_NUM_HS_THREADS(num_hs_threads); tmp |= S_008C1C_NUM_LS_THREADS(num_ls_threads); - r600_pipe_state_add_reg(rstate, R_008C1C_SQ_THREAD_RESOURCE_MGMT_2, tmp, NULL, 0); + r600_store_value(cb, tmp); /* R_008C1C_SQ_THREAD_RESOURCE_MGMT_2 */ - tmp = 0; - tmp |= S_008C20_NUM_PS_STACK_ENTRIES(num_ps_stack_entries); + tmp = S_008C20_NUM_PS_STACK_ENTRIES(num_ps_stack_entries); tmp |= S_008C20_NUM_VS_STACK_ENTRIES(num_vs_stack_entries); - r600_pipe_state_add_reg(rstate, R_008C20_SQ_STACK_RESOURCE_MGMT_1, tmp, NULL, 0); + r600_store_value(cb, tmp); /* R_008C20_SQ_STACK_RESOURCE_MGMT_1 */ - tmp = 0; - tmp |= S_008C24_NUM_GS_STACK_ENTRIES(num_gs_stack_entries); + tmp = S_008C24_NUM_GS_STACK_ENTRIES(num_gs_stack_entries); tmp |= S_008C24_NUM_ES_STACK_ENTRIES(num_es_stack_entries); - r600_pipe_state_add_reg(rstate, R_008C24_SQ_STACK_RESOURCE_MGMT_2, tmp, NULL, 0); + r600_store_value(cb, tmp); /* R_008C24_SQ_STACK_RESOURCE_MGMT_2 */ - tmp = 0; - tmp |= S_008C28_NUM_HS_STACK_ENTRIES(num_hs_stack_entries); + tmp = S_008C28_NUM_HS_STACK_ENTRIES(num_hs_stack_entries); tmp |= S_008C28_NUM_LS_STACK_ENTRIES(num_ls_stack_entries); - r600_pipe_state_add_reg(rstate, R_008C28_SQ_STACK_RESOURCE_MGMT_3, tmp, NULL, 0); - - tmp = 0; - tmp |= S_008E2C_NUM_PS_LDS(0x1000); - tmp |= S_008E2C_NUM_LS_LDS(0x1000); - r600_pipe_state_add_reg(rstate, R_008E2C_SQ_LDS_RESOURCE_MGMT, tmp, NULL, 0); - - r600_pipe_state_add_reg(rstate, R_009100_SPI_CONFIG_CNTL, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_00913C_SPI_CONFIG_CNTL_1, S_00913C_VTX_DONE_DELAY(4), NULL, 0); - -#if 0 - r600_pipe_state_add_reg(rstate, R_028350_SX_MISC, 0x0, NULL, 0); - - r600_pipe_state_add_reg(rstate, R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, 0x0, NULL, 0); -#endif - r600_pipe_state_add_reg(rstate, R_028A4C_PA_SC_MODE_CNTL_1, 0x0, NULL, 0); - - r600_pipe_state_add_reg(rstate, R_028900_SQ_ESGS_RING_ITEMSIZE, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028904_SQ_GSVS_RING_ITEMSIZE, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028908_SQ_ESTMP_RING_ITEMSIZE, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_02890C_SQ_GSTMP_RING_ITEMSIZE, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028910_SQ_VSTMP_RING_ITEMSIZE, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028914_SQ_PSTMP_RING_ITEMSIZE, 0x0, NULL, 0); - - r600_pipe_state_add_reg(rstate, R_02891C_SQ_GS_VERT_ITEMSIZE, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028920_SQ_GS_VERT_ITEMSIZE_1, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028924_SQ_GS_VERT_ITEMSIZE_2, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028928_SQ_GS_VERT_ITEMSIZE_3, 0x0, NULL, 0); - - r600_pipe_state_add_reg(rstate, R_028A10_VGT_OUTPUT_PATH_CNTL, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A14_VGT_HOS_CNTL, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A18_VGT_HOS_MAX_TESS_LEVEL, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A1C_VGT_HOS_MIN_TESS_LEVEL, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A20_VGT_HOS_REUSE_DEPTH, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A24_VGT_GROUP_PRIM_TYPE, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A28_VGT_GROUP_FIRST_DECR, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A2C_VGT_GROUP_DECR, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A30_VGT_GROUP_VECT_0_CNTL, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A34_VGT_GROUP_VECT_1_CNTL, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A38_VGT_GROUP_VECT_0_FMT_CNTL, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A3C_VGT_GROUP_VECT_1_FMT_CNTL, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A40_VGT_GS_MODE, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028B94_VGT_STRMOUT_CONFIG, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028B98_VGT_STRMOUT_BUFFER_CONFIG, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028AB4_VGT_REUSE_OFF, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028AB8_VGT_VTX_CNT_EN, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_008A14_PA_CL_ENHANCE, (3 << 1) | 1, NULL, 0); - - r600_pipe_state_add_reg(rstate, R_028380_SQ_VTX_SEMANTIC_0, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028384_SQ_VTX_SEMANTIC_1, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028388_SQ_VTX_SEMANTIC_2, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_02838C_SQ_VTX_SEMANTIC_3, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028390_SQ_VTX_SEMANTIC_4, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028394_SQ_VTX_SEMANTIC_5, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028398_SQ_VTX_SEMANTIC_6, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_02839C_SQ_VTX_SEMANTIC_7, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283A0_SQ_VTX_SEMANTIC_8, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283A4_SQ_VTX_SEMANTIC_9, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283A8_SQ_VTX_SEMANTIC_10, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283AC_SQ_VTX_SEMANTIC_11, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283B0_SQ_VTX_SEMANTIC_12, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283B4_SQ_VTX_SEMANTIC_13, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283B8_SQ_VTX_SEMANTIC_14, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283BC_SQ_VTX_SEMANTIC_15, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283C0_SQ_VTX_SEMANTIC_16, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283C4_SQ_VTX_SEMANTIC_17, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283C8_SQ_VTX_SEMANTIC_18, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283CC_SQ_VTX_SEMANTIC_19, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283D0_SQ_VTX_SEMANTIC_20, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283D4_SQ_VTX_SEMANTIC_21, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283D8_SQ_VTX_SEMANTIC_22, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283DC_SQ_VTX_SEMANTIC_23, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283E0_SQ_VTX_SEMANTIC_24, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283E4_SQ_VTX_SEMANTIC_25, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283E8_SQ_VTX_SEMANTIC_26, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283EC_SQ_VTX_SEMANTIC_27, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283F0_SQ_VTX_SEMANTIC_28, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283F4_SQ_VTX_SEMANTIC_29, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283F8_SQ_VTX_SEMANTIC_30, 0x0, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0283FC_SQ_VTX_SEMANTIC_31, 0x0, NULL, 0); - - r600_pipe_state_add_reg(rstate, R_028810_PA_CL_CLIP_CNTL, 0x0, NULL, 0); - - r600_context_pipe_state_set(rctx, rstate); + r600_store_value(cb, tmp); /* R_008C28_SQ_STACK_RESOURCE_MGMT_3 */ + + r600_store_config_reg(cb, R_008E2C_SQ_LDS_RESOURCE_MGMT, + S_008E2C_NUM_PS_LDS(0x1000) | S_008E2C_NUM_LS_LDS(0x1000)); + + r600_store_config_reg(cb, R_009100_SPI_CONFIG_CNTL, 0); + r600_store_config_reg(cb, R_00913C_SPI_CONFIG_CNTL_1, S_00913C_VTX_DONE_DELAY(4)); + + r600_store_context_reg(cb, R_028A4C_PA_SC_MODE_CNTL_1, 0); + + r600_store_context_reg_seq(cb, R_028900_SQ_ESGS_RING_ITEMSIZE, 6); + r600_store_value(cb, 0); /* R_028900_SQ_ESGS_RING_ITEMSIZE */ + r600_store_value(cb, 0); /* R_028904_SQ_GSVS_RING_ITEMSIZE */ + r600_store_value(cb, 0); /* R_028908_SQ_ESTMP_RING_ITEMSIZE */ + r600_store_value(cb, 0); /* R_02890C_SQ_GSTMP_RING_ITEMSIZE */ + r600_store_value(cb, 0); /* R_028910_SQ_VSTMP_RING_ITEMSIZE */ + r600_store_value(cb, 0); /* R_028914_SQ_PSTMP_RING_ITEMSIZE */ + + r600_store_context_reg_seq(cb, R_02891C_SQ_GS_VERT_ITEMSIZE, 4); + r600_store_value(cb, 0); /* R_02891C_SQ_GS_VERT_ITEMSIZE */ + r600_store_value(cb, 0); /* R_028920_SQ_GS_VERT_ITEMSIZE_1 */ + r600_store_value(cb, 0); /* R_028924_SQ_GS_VERT_ITEMSIZE_2 */ + r600_store_value(cb, 0); /* R_028928_SQ_GS_VERT_ITEMSIZE_3 */ + + r600_store_context_reg_seq(cb, R_028A10_VGT_OUTPUT_PATH_CNTL, 13); + r600_store_value(cb, 0); /* R_028A10_VGT_OUTPUT_PATH_CNTL */ + r600_store_value(cb, 0); /* R_028A14_VGT_HOS_CNTL */ + r600_store_value(cb, 0); /* R_028A18_VGT_HOS_MAX_TESS_LEVEL */ + r600_store_value(cb, 0); /* R_028A1C_VGT_HOS_MIN_TESS_LEVEL */ + r600_store_value(cb, 0); /* R_028A20_VGT_HOS_REUSE_DEPTH */ + r600_store_value(cb, 0); /* R_028A24_VGT_GROUP_PRIM_TYPE */ + r600_store_value(cb, 0); /* R_028A28_VGT_GROUP_FIRST_DECR */ + r600_store_value(cb, 0); /* R_028A2C_VGT_GROUP_DECR */ + r600_store_value(cb, 0); /* R_028A30_VGT_GROUP_VECT_0_CNTL */ + r600_store_value(cb, 0); /* R_028A34_VGT_GROUP_VECT_1_CNTL */ + r600_store_value(cb, 0); /* R_028A38_VGT_GROUP_VECT_0_FMT_CNTL */ + r600_store_value(cb, 0); /* R_028A3C_VGT_GROUP_VECT_1_FMT_CNTL */ + r600_store_value(cb, 0); /* R_028A40_VGT_GS_MODE */ + + r600_store_context_reg_seq(cb, R_028B94_VGT_STRMOUT_CONFIG, 2); + r600_store_value(cb, 0); /* R_028B94_VGT_STRMOUT_CONFIG */ + r600_store_value(cb, 0); /* R_028B98_VGT_STRMOUT_BUFFER_CONFIG */ + + r600_store_context_reg_seq(cb, R_028AB4_VGT_REUSE_OFF, 2); + r600_store_value(cb, 0); /* R_028AB4_VGT_REUSE_OFF */ + r600_store_value(cb, 0); /* R_028AB8_VGT_VTX_CNT_EN */ + + r600_store_config_reg(cb, R_008A14_PA_CL_ENHANCE, (3 << 1) | 1); + + r600_store_context_reg_seq(cb, R_028380_SQ_VTX_SEMANTIC_0, 32); + r600_store_value(cb, 0); /* R_028380_SQ_VTX_SEMANTIC_0 */ + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); + r600_store_value(cb, 0); /* R_028380_SQ_VTX_SEMANTIC_31 */ } void evergreen_polygon_offset_update(struct r600_context *rctx) diff --git a/src/gallium/drivers/r600/evergreend.h b/src/gallium/drivers/r600/evergreend.h index d1dbd843a28..66b3d1fb9b4 100644 --- a/src/gallium/drivers/r600/evergreend.h +++ b/src/gallium/drivers/r600/evergreend.h @@ -117,7 +117,6 @@ #define PKT3_IT_OPCODE_C 0xFFFF00FF #define PKT3_PREDICATE(x) (((x) >> 0) & 0x1) #define PKT0(index, count) (PKT_TYPE_S(0) | PKT0_BASE_INDEX_S(index) | PKT_COUNT_S(count)) -#define PKT3(op, count, predicate) (PKT_TYPE_S(3) | PKT3_IT_OPCODE_S(op) | PKT_COUNT_S(count) | PKT3_PREDICATE(predicate)) /* Registers */ #define R_0084FC_CP_STRMOUT_CNTL 0x000084FC diff --git a/src/gallium/drivers/r600/r600_hw_context.c b/src/gallium/drivers/r600/r600_hw_context.c index 1de345c92d1..ef29ddf51a1 100644 --- a/src/gallium/drivers/r600/r600_hw_context.c +++ b/src/gallium/drivers/r600/r600_hw_context.c @@ -127,23 +127,6 @@ static inline void r600_context_ps_partial_flush(struct r600_context *ctx) ctx->flags &= ~R600_CONTEXT_DRAW_PENDING; } -void r600_init_cs(struct r600_context *ctx) -{ - struct radeon_winsys_cs *cs = ctx->cs; - - /* R6xx requires this packet at the start of each command buffer */ - if (ctx->family < CHIP_RV770) { - cs->buf[cs->cdw++] = PKT3(PKT3_START_3D_CMDBUF, 0, 0); - cs->buf[cs->cdw++] = 0x00000000; - } - /* All asics require this one */ - cs->buf[cs->cdw++] = PKT3(PKT3_CONTEXT_CONTROL, 1, 0); - cs->buf[cs->cdw++] = 0x80000000; - cs->buf[cs->cdw++] = 0x80000000; - - ctx->init_dwords = cs->cdw; -} - static void r600_init_block(struct r600_context *ctx, struct r600_block *block, const struct r600_reg *reg, int index, int nreg, @@ -262,17 +245,8 @@ int r600_context_add_block(struct r600_context *ctx, const struct r600_reg *reg, /* R600/R700 configuration */ static const struct r600_reg r600_config_reg_list[] = { {R_008958_VGT_PRIMITIVE_TYPE, 0, 0}, - {R_008C00_SQ_CONFIG, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, {R_008C04_SQ_GPR_RESOURCE_MGMT_1, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_008C08_SQ_GPR_RESOURCE_MGMT_2, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_008C0C_SQ_THREAD_RESOURCE_MGMT, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_008C10_SQ_STACK_RESOURCE_MGMT_1, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_008C14_SQ_STACK_RESOURCE_MGMT_2, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, {R_009508_TA_CNTL_AUX, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_009714_VC_ENHANCE, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_009830_DB_DEBUG, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, - {R_009838_DB_WATERMARKS, REG_FLAG_ENABLE_ALWAYS | REG_FLAG_FLUSH_CHANGE, 0}, }; static const struct r600_reg r600_ctl_const_list[] = { @@ -281,35 +255,7 @@ static const struct r600_reg r600_ctl_const_list[] = { }; static const struct r600_reg r600_context_reg_list[] = { - {R_028350_SX_MISC, 0, 0}, - {R_0286C8_SPI_THREAD_GROUPING, 0, 0}, - {R_0288A8_SQ_ESGS_RING_ITEMSIZE, 0, 0}, - {R_0288AC_SQ_GSVS_RING_ITEMSIZE, 0, 0}, - {R_0288B0_SQ_ESTMP_RING_ITEMSIZE, 0, 0}, - {R_0288B4_SQ_GSTMP_RING_ITEMSIZE, 0, 0}, - {R_0288B8_SQ_VSTMP_RING_ITEMSIZE, 0, 0}, - {R_0288BC_SQ_PSTMP_RING_ITEMSIZE, 0, 0}, - {R_0288C0_SQ_FBUF_RING_ITEMSIZE, 0, 0}, - {R_0288C4_SQ_REDUC_RING_ITEMSIZE, 0, 0}, - {R_0288C8_SQ_GS_VERT_ITEMSIZE, 0, 0}, - {R_028A10_VGT_OUTPUT_PATH_CNTL, 0, 0}, - {R_028A14_VGT_HOS_CNTL, 0, 0}, - {R_028A18_VGT_HOS_MAX_TESS_LEVEL, 0, 0}, - {R_028A1C_VGT_HOS_MIN_TESS_LEVEL, 0, 0}, - {R_028A20_VGT_HOS_REUSE_DEPTH, 0, 0}, - {R_028A24_VGT_GROUP_PRIM_TYPE, 0, 0}, - {R_028A28_VGT_GROUP_FIRST_DECR, 0, 0}, - {R_028A2C_VGT_GROUP_DECR, 0, 0}, - {R_028A30_VGT_GROUP_VECT_0_CNTL, 0, 0}, - {R_028A34_VGT_GROUP_VECT_1_CNTL, 0, 0}, - {R_028A38_VGT_GROUP_VECT_0_FMT_CNTL, 0, 0}, - {R_028A3C_VGT_GROUP_VECT_1_FMT_CNTL, 0, 0}, - {R_028A40_VGT_GS_MODE, 0, 0}, {R_028A4C_PA_SC_MODE_CNTL, 0, 0}, - {R_028AB0_VGT_STRMOUT_EN, 0, 0}, - {R_028AB4_VGT_REUSE_OFF, 0, 0}, - {R_028AB8_VGT_VTX_CNT_EN, 0, 0}, - {R_028B20_VGT_STRMOUT_BUFFER_EN, 0, 0}, {R_028028_DB_STENCIL_CLEAR, 0, 0}, {R_02802C_DB_DEPTH_CLEAR, 0, 0}, {GROUP_FORCE_NEW_BLOCK, 0, 0}, @@ -629,10 +575,7 @@ static const struct r600_reg r600_context_reg_list[] = { {R_028404_VGT_MIN_VTX_INDX, 0, 0}, {R_028408_VGT_INDX_OFFSET, 0, 0}, {R_02840C_VGT_MULTI_PRIM_IB_RESET_INDX, 0, 0}, - {R_028A84_VGT_PRIMITIVEID_EN, 0, 0}, {R_028A94_VGT_MULTI_PRIM_IB_RESET_EN, 0, 0}, - {R_028AA0_VGT_INSTANCE_STEP_RATE_0, 0, 0}, - {R_028AA4_VGT_INSTANCE_STEP_RATE_1, 0, 0}, }; /* SHADER RESOURCE R600/R700 */ @@ -898,8 +841,8 @@ int r600_context_init(struct r600_context *ctx) goto out_err; ctx->cs = ctx->ws->cs_create(ctx->ws); + r600_emit_atom(ctx, &ctx->atom_start_cs.atom); - r600_init_cs(ctx); ctx->max_db = 4; return 0; out_err: @@ -1419,7 +1362,7 @@ void r600_context_flush(struct r600_context *ctx, unsigned flags) bool queries_suspended = false; bool streamout_suspended = false; - if (cs->cdw == ctx->init_dwords) + if (cs->cdw == ctx->atom_start_cs.atom.num_dw) return; /* suspend queries */ @@ -1448,7 +1391,7 @@ void r600_context_flush(struct r600_context *ctx, unsigned flags) ctx->pm4_dirty_cdwords = 0; ctx->flags = 0; - r600_init_cs(ctx); + r600_emit_atom(ctx, &ctx->atom_start_cs.atom); if (streamout_suspended) { ctx->streamout_start = TRUE; diff --git a/src/gallium/drivers/r600/r600_hw_context_priv.h b/src/gallium/drivers/r600/r600_hw_context_priv.h index 278bc0d6fab..1e5713873d9 100644 --- a/src/gallium/drivers/r600/r600_hw_context_priv.h +++ b/src/gallium/drivers/r600/r600_hw_context_priv.h @@ -60,7 +60,6 @@ void r600_context_pipe_state_set_resource(struct r600_context *ctx, struct r600_ void r600_context_dirty_block(struct r600_context *ctx, struct r600_block *block, int dirty, int index); int r600_setup_block_table(struct r600_context *ctx); -void r600_init_cs(struct r600_context *ctx); int r600_resource_init(struct r600_context *ctx, struct r600_range *range, unsigned offset, unsigned nblocks, unsigned stride, struct r600_reg *reg, int nreg, unsigned offset_base); /* diff --git a/src/gallium/drivers/r600/r600_pipe.c b/src/gallium/drivers/r600/r600_pipe.c index 2369ed4cf20..c8ee331789b 100644 --- a/src/gallium/drivers/r600/r600_pipe.c +++ b/src/gallium/drivers/r600/r600_pipe.c @@ -207,6 +207,7 @@ static void r600_destroy_context(struct pipe_context *context) r600_update_num_contexts(rctx->screen, -1); + r600_release_command_buffer(&rctx->atom_start_cs); FREE(rctx); } @@ -246,21 +247,21 @@ static struct pipe_context *r600_create_context(struct pipe_screen *screen, void case R600: case R700: r600_init_state_functions(rctx); + r600_init_atom_start_cs(rctx); if (r600_context_init(rctx)) { r600_destroy_context(&rctx->context); return NULL; } - r600_init_config(rctx); rctx->custom_dsa_flush = r600_create_db_flush_dsa(rctx); break; case EVERGREEN: case CAYMAN: evergreen_init_state_functions(rctx); + evergreen_init_atom_start_cs(rctx); if (evergreen_context_init(rctx)) { r600_destroy_context(&rctx->context); return NULL; } - evergreen_init_config(rctx); rctx->custom_dsa_flush = evergreen_create_db_flush_dsa(rctx); break; default: diff --git a/src/gallium/drivers/r600/r600_pipe.h b/src/gallium/drivers/r600/r600_pipe.h index 8eaa988fb38..8889ba7430f 100644 --- a/src/gallium/drivers/r600/r600_pipe.h +++ b/src/gallium/drivers/r600/r600_pipe.h @@ -67,6 +67,14 @@ struct r600_atom { struct list_head head; }; +/* This is an atom containing GPU commands that never change. + * This is supposed to be copied directly into the CS. */ +struct r600_command_buffer { + struct r600_atom atom; + uint32_t *buf; + unsigned max_num_dw; +}; + struct r600_atom_surface_sync { struct r600_atom atom; unsigned flush_flags; /* CP_COHER_CNTL */ @@ -245,7 +253,6 @@ struct r600_context { struct pipe_stencil_ref stencil_ref; struct pipe_viewport_state viewport; struct pipe_clip_state clip; - struct r600_pipe_state config; struct r600_pipe_shader *ps_shader; struct r600_pipe_shader *vs_shader; struct r600_pipe_state vs_const_buffer; @@ -277,6 +284,7 @@ struct r600_context { /* States based on r600_state. */ struct list_head dirty_states; + struct r600_command_buffer atom_start_cs; /* invariant state mostly */ struct r600_atom_surface_sync atom_surface_sync; struct r600_atom atom_r6xx_flush_and_inv; @@ -292,7 +300,6 @@ struct r600_context { struct list_head enable_list; unsigned pm4_dirty_cdwords; unsigned ctx_pm4_ndwords; - unsigned init_dwords; /* The list of active queries. Only one query of each type can be active. */ struct list_head active_query_list; @@ -336,7 +343,7 @@ static INLINE void r600_atom_dirty(struct r600_context *rctx, struct r600_atom * /* evergreen_state.c */ void evergreen_init_state_functions(struct r600_context *rctx); -void evergreen_init_config(struct r600_context *rctx); +void evergreen_init_atom_start_cs(struct r600_context *rctx); void evergreen_pipe_shader_ps(struct pipe_context *ctx, struct r600_pipe_shader *shader); void evergreen_pipe_shader_vs(struct pipe_context *ctx, struct r600_pipe_shader *shader); void evergreen_fetch_shader(struct pipe_context *ctx, struct r600_vertex_element *ve); @@ -394,7 +401,7 @@ int r600_find_vs_semantic_index(struct r600_shader *vs, /* r600_state.c */ void r600_update_sampler_states(struct r600_context *rctx); void r600_init_state_functions(struct r600_context *rctx); -void r600_init_config(struct r600_context *rctx); +void r600_init_atom_start_cs(struct r600_context *rctx); void r600_pipe_shader_ps(struct pipe_context *ctx, struct r600_pipe_shader *shader); void r600_pipe_shader_vs(struct pipe_context *ctx, struct r600_pipe_shader *shader); void r600_fetch_shader(struct pipe_context *ctx, struct r600_vertex_element *ve); @@ -478,6 +485,58 @@ unsigned r600_tex_filter(unsigned filter); unsigned r600_tex_mipfilter(unsigned filter); unsigned r600_tex_compare(unsigned compare); +/* + * Helpers for building command buffers + */ + +#define PKT3_SET_CONFIG_REG 0x68 +#define PKT3_SET_CONTEXT_REG 0x69 + +#define R600_CONFIG_REG_OFFSET 0x8000 +#define R600_CONTEXT_REG_OFFSET 0x28000 + +#define PKT_TYPE_S(x) (((x) & 0x3) << 30) +#define PKT_COUNT_S(x) (((x) & 0x3FFF) << 16) +#define PKT3_IT_OPCODE_S(x) (((x) & 0xFF) << 8) +#define PKT3_PREDICATE(x) (((x) >> 0) & 0x1) +#define PKT3(op, count, predicate) (PKT_TYPE_S(3) | PKT_COUNT_S(count) | PKT3_IT_OPCODE_S(op) | PKT3_PREDICATE(predicate)) + +static INLINE void r600_store_value(struct r600_command_buffer *cb, unsigned value) +{ + cb->buf[cb->atom.num_dw++] = value; +} + +static INLINE void r600_store_config_reg_seq(struct r600_command_buffer *cb, unsigned reg, unsigned num) +{ + assert(reg < R600_CONTEXT_REG_OFFSET); + assert(cb->atom.num_dw+2+num <= cb->max_num_dw); + cb->buf[cb->atom.num_dw++] = PKT3(PKT3_SET_CONFIG_REG, num, 0); + cb->buf[cb->atom.num_dw++] = (reg - R600_CONFIG_REG_OFFSET) >> 2; +} + +static INLINE void r600_store_context_reg_seq(struct r600_command_buffer *cb, unsigned reg, unsigned num) +{ + assert(reg >= R600_CONTEXT_REG_OFFSET); + assert(cb->atom.num_dw+2+num <= cb->max_num_dw); + cb->buf[cb->atom.num_dw++] = PKT3(PKT3_SET_CONTEXT_REG, num, 0); + cb->buf[cb->atom.num_dw++] = (reg - R600_CONTEXT_REG_OFFSET) >> 2; +} + +static INLINE void r600_store_config_reg(struct r600_command_buffer *cb, unsigned reg, unsigned value) +{ + r600_store_config_reg_seq(cb, reg, 1); + r600_store_value(cb, value); +} + +static INLINE void r600_store_context_reg(struct r600_command_buffer *cb, unsigned reg, unsigned value) +{ + r600_store_context_reg_seq(cb, reg, 1); + r600_store_value(cb, value); +} + +void r600_init_command_buffer(struct r600_command_buffer *cb, unsigned num_dw, enum r600_atom_flags flags); +void r600_release_command_buffer(struct r600_command_buffer *cb); + /* * common helpers */ diff --git a/src/gallium/drivers/r600/r600_state.c b/src/gallium/drivers/r600/r600_state.c index bbba9a2d11f..939bbe4126b 100644 --- a/src/gallium/drivers/r600/r600_state.c +++ b/src/gallium/drivers/r600/r600_state.c @@ -1787,7 +1787,7 @@ void r600_adjust_gprs(struct r600_context *rctx) r600_context_pipe_state_set(rctx, &rstate); } -void r600_init_config(struct r600_context *rctx) +void r600_init_atom_start_cs(struct r600_context *rctx) { int ps_prio; int vs_prio; @@ -1807,9 +1807,21 @@ void r600_init_config(struct r600_context *rctx) int num_gs_stack_entries; int num_es_stack_entries; enum radeon_family family; - struct r600_pipe_state *rstate = &rctx->config; + struct r600_command_buffer *cb = &rctx->atom_start_cs; uint32_t tmp; + r600_init_command_buffer(cb, 256, EMIT_EARLY); + + /* R6xx requires this packet at the start of each command buffer */ + if (rctx->chip_class == R600) { + r600_store_value(cb, PKT3(PKT3_START_3D_CMDBUF, 0, 0)); + r600_store_value(cb, 0); + } + /* All asics require this one */ + r600_store_value(cb, PKT3(PKT3_CONTEXT_CONTROL, 1, 0)); + r600_store_value(cb, 0x80000000); + r600_store_value(cb, 0x80000000); + family = rctx->family; ps_prio = 0; vs_prio = 1; @@ -1931,8 +1943,7 @@ void r600_init_config(struct r600_context *rctx) rctx->default_ps_gprs = num_ps_gprs; rctx->default_vs_gprs = num_vs_gprs; - - rstate->id = R600_PIPE_STATE_CONFIG; + rctx->r6xx_num_clause_temp_gprs = num_temp_gprs; /* SQ_CONFIG */ tmp = 0; @@ -1953,91 +1964,82 @@ void r600_init_config(struct r600_context *rctx) tmp |= S_008C00_VS_PRIO(vs_prio); tmp |= S_008C00_GS_PRIO(gs_prio); tmp |= S_008C00_ES_PRIO(es_prio); - r600_pipe_state_add_reg(rstate, R_008C00_SQ_CONFIG, tmp, NULL, 0); - - /* SQ_GPR_RESOURCE_MGMT_1 */ - tmp = 0; - tmp |= S_008C04_NUM_PS_GPRS(num_ps_gprs); - tmp |= S_008C04_NUM_VS_GPRS(num_vs_gprs); - tmp |= S_008C04_NUM_CLAUSE_TEMP_GPRS(num_temp_gprs); - rctx->r6xx_num_clause_temp_gprs = num_temp_gprs; - r600_pipe_state_add_reg(rstate, R_008C04_SQ_GPR_RESOURCE_MGMT_1, tmp, NULL, 0); + r600_store_config_reg(cb, R_008C00_SQ_CONFIG, tmp); /* SQ_GPR_RESOURCE_MGMT_2 */ - tmp = 0; - tmp |= S_008C08_NUM_GS_GPRS(num_gs_gprs); + tmp = S_008C08_NUM_GS_GPRS(num_gs_gprs); tmp |= S_008C08_NUM_ES_GPRS(num_es_gprs); - r600_pipe_state_add_reg(rstate, R_008C08_SQ_GPR_RESOURCE_MGMT_2, tmp, NULL, 0); + r600_store_config_reg_seq(cb, R_008C08_SQ_GPR_RESOURCE_MGMT_2, 4); + r600_store_value(cb, tmp); /* SQ_THREAD_RESOURCE_MGMT */ - tmp = 0; - tmp |= S_008C0C_NUM_PS_THREADS(num_ps_threads); + tmp = S_008C0C_NUM_PS_THREADS(num_ps_threads); tmp |= S_008C0C_NUM_VS_THREADS(num_vs_threads); tmp |= S_008C0C_NUM_GS_THREADS(num_gs_threads); tmp |= S_008C0C_NUM_ES_THREADS(num_es_threads); - r600_pipe_state_add_reg(rstate, R_008C0C_SQ_THREAD_RESOURCE_MGMT, tmp, NULL, 0); + r600_store_value(cb, tmp); /* R_008C0C_SQ_THREAD_RESOURCE_MGMT */ /* SQ_STACK_RESOURCE_MGMT_1 */ - tmp = 0; - tmp |= S_008C10_NUM_PS_STACK_ENTRIES(num_ps_stack_entries); + tmp = S_008C10_NUM_PS_STACK_ENTRIES(num_ps_stack_entries); tmp |= S_008C10_NUM_VS_STACK_ENTRIES(num_vs_stack_entries); - r600_pipe_state_add_reg(rstate, R_008C10_SQ_STACK_RESOURCE_MGMT_1, tmp, NULL, 0); + r600_store_value(cb, tmp); /* R_008C10_SQ_STACK_RESOURCE_MGMT_1 */ /* SQ_STACK_RESOURCE_MGMT_2 */ - tmp = 0; - tmp |= S_008C14_NUM_GS_STACK_ENTRIES(num_gs_stack_entries); + tmp = S_008C14_NUM_GS_STACK_ENTRIES(num_gs_stack_entries); tmp |= S_008C14_NUM_ES_STACK_ENTRIES(num_es_stack_entries); - r600_pipe_state_add_reg(rstate, R_008C14_SQ_STACK_RESOURCE_MGMT_2, tmp, NULL, 0); + r600_store_value(cb, tmp); /* R_008C14_SQ_STACK_RESOURCE_MGMT_2 */ + + r600_store_config_reg(cb, R_009714_VC_ENHANCE, 0); - r600_pipe_state_add_reg(rstate, R_009714_VC_ENHANCE, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028350_SX_MISC, 0x00000000, NULL, 0); + r600_store_context_reg(cb, R_028350_SX_MISC, 0); if (rctx->chip_class >= R700) { - r600_pipe_state_add_reg(rstate, R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, 0x00004000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_009830_DB_DEBUG, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_009838_DB_WATERMARKS, 0x00420204, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0286C8_SPI_THREAD_GROUPING, 0x00000000, NULL, 0); + r600_store_config_reg(cb, R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, 0x00004000); + r600_store_config_reg(cb, R_009830_DB_DEBUG, 0); + r600_store_config_reg(cb, R_009838_DB_WATERMARKS, 0x00420204); + r600_store_context_reg(cb, R_0286C8_SPI_THREAD_GROUPING, 0); } else { - r600_pipe_state_add_reg(rstate, R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_009830_DB_DEBUG, 0x82000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_009838_DB_WATERMARKS, 0x01020204, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0286C8_SPI_THREAD_GROUPING, 0x00000001, NULL, 0); + r600_store_config_reg(cb, R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, 0); + r600_store_config_reg(cb, R_009830_DB_DEBUG, 0x82000000); + r600_store_config_reg(cb, R_009838_DB_WATERMARKS, 0x01020204); + r600_store_context_reg(cb, R_0286C8_SPI_THREAD_GROUPING, 1); } - r600_pipe_state_add_reg(rstate, R_0288A8_SQ_ESGS_RING_ITEMSIZE, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0288AC_SQ_GSVS_RING_ITEMSIZE, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0288B0_SQ_ESTMP_RING_ITEMSIZE, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0288B4_SQ_GSTMP_RING_ITEMSIZE, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0288B8_SQ_VSTMP_RING_ITEMSIZE, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0288BC_SQ_PSTMP_RING_ITEMSIZE, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0288C0_SQ_FBUF_RING_ITEMSIZE, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0288C4_SQ_REDUC_RING_ITEMSIZE, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_0288C8_SQ_GS_VERT_ITEMSIZE, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A10_VGT_OUTPUT_PATH_CNTL, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A14_VGT_HOS_CNTL, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A18_VGT_HOS_MAX_TESS_LEVEL, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A1C_VGT_HOS_MIN_TESS_LEVEL, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A20_VGT_HOS_REUSE_DEPTH, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A24_VGT_GROUP_PRIM_TYPE, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A28_VGT_GROUP_FIRST_DECR, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A2C_VGT_GROUP_DECR, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A30_VGT_GROUP_VECT_0_CNTL, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A34_VGT_GROUP_VECT_1_CNTL, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A38_VGT_GROUP_VECT_0_FMT_CNTL, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A3C_VGT_GROUP_VECT_1_FMT_CNTL, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A40_VGT_GS_MODE, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028AB0_VGT_STRMOUT_EN, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028AB4_VGT_REUSE_OFF, 0x00000001, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028AB8_VGT_VTX_CNT_EN, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028B20_VGT_STRMOUT_BUFFER_EN, 0x00000000, NULL, 0); - - r600_pipe_state_add_reg(rstate, R_02840C_VGT_MULTI_PRIM_IB_RESET_INDX, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A84_VGT_PRIMITIVEID_EN, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A94_VGT_MULTI_PRIM_IB_RESET_EN, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028AA0_VGT_INSTANCE_STEP_RATE_0, 0x00000000, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028AA4_VGT_INSTANCE_STEP_RATE_1, 0x00000000, NULL, 0); - r600_context_pipe_state_set(rctx, rstate); - - r600_set_seamless_cubemap(rctx, FALSE); + r600_store_context_reg_seq(cb, R_0288A8_SQ_ESGS_RING_ITEMSIZE, 9); + r600_store_value(cb, 0); /* R_0288A8_SQ_ESGS_RING_ITEMSIZE */ + r600_store_value(cb, 0); /* R_0288AC_SQ_GSVS_RING_ITEMSIZE */ + r600_store_value(cb, 0); /* R_0288B0_SQ_ESTMP_RING_ITEMSIZE */ + r600_store_value(cb, 0); /* R_0288B4_SQ_GSTMP_RING_ITEMSIZE */ + r600_store_value(cb, 0); /* R_0288B8_SQ_VSTMP_RING_ITEMSIZE */ + r600_store_value(cb, 0); /* R_0288BC_SQ_PSTMP_RING_ITEMSIZE */ + r600_store_value(cb, 0); /* R_0288C0_SQ_FBUF_RING_ITEMSIZE */ + r600_store_value(cb, 0); /* R_0288C4_SQ_REDUC_RING_ITEMSIZE */ + r600_store_value(cb, 0); /* R_0288C8_SQ_GS_VERT_ITEMSIZE */ + + r600_store_context_reg_seq(cb, R_028A10_VGT_OUTPUT_PATH_CNTL, 13); + r600_store_value(cb, 0); /* R_028A10_VGT_OUTPUT_PATH_CNTL */ + r600_store_value(cb, 0); /* R_028A14_VGT_HOS_CNTL */ + r600_store_value(cb, 0); /* R_028A18_VGT_HOS_MAX_TESS_LEVEL */ + r600_store_value(cb, 0); /* R_028A1C_VGT_HOS_MIN_TESS_LEVEL */ + r600_store_value(cb, 0); /* R_028A20_VGT_HOS_REUSE_DEPTH */ + r600_store_value(cb, 0); /* R_028A24_VGT_GROUP_PRIM_TYPE */ + r600_store_value(cb, 0); /* R_028A28_VGT_GROUP_FIRST_DECR */ + r600_store_value(cb, 0); /* R_028A2C_VGT_GROUP_DECR */ + r600_store_value(cb, 0); /* R_028A30_VGT_GROUP_VECT_0_CNTL */ + r600_store_value(cb, 0); /* R_028A34_VGT_GROUP_VECT_1_CNTL */ + r600_store_value(cb, 0); /* R_028A38_VGT_GROUP_VECT_0_FMT_CNTL */ + r600_store_value(cb, 0); /* R_028A3C_VGT_GROUP_VECT_1_FMT_CNTL */ + r600_store_value(cb, 0); /* R_028A40_VGT_GS_MODE, 0); */ + + r600_store_context_reg(cb, R_028A84_VGT_PRIMITIVEID_EN, 0); + r600_store_context_reg(cb, R_028AA0_VGT_INSTANCE_STEP_RATE_0, 0); + r600_store_context_reg(cb, R_028AA4_VGT_INSTANCE_STEP_RATE_1, 0); + + r600_store_context_reg_seq(cb, R_028AB0_VGT_STRMOUT_EN, 3); + r600_store_value(cb, 0); /* R_028AB0_VGT_STRMOUT_EN */ + r600_store_value(cb, 1); /* R_028AB4_VGT_REUSE_OFF */ + r600_store_value(cb, 0); /* R_028AB8_VGT_VTX_CNT_EN */ + + r600_store_context_reg(cb, R_028B20_VGT_STRMOUT_BUFFER_EN, 0); } void r600_pipe_shader_ps(struct pipe_context *ctx, struct r600_pipe_shader *shader) diff --git a/src/gallium/drivers/r600/r600_state_common.c b/src/gallium/drivers/r600/r600_state_common.c index 36b3d53fe56..031f2000c31 100644 --- a/src/gallium/drivers/r600/r600_state_common.c +++ b/src/gallium/drivers/r600/r600_state_common.c @@ -34,6 +34,30 @@ #include "r600_pipe.h" #include "r600d.h" +static void r600_emit_command_buffer(struct r600_context *rctx, struct r600_atom *atom) +{ + struct radeon_winsys_cs *cs = rctx->cs; + struct r600_command_buffer *cb = (struct r600_command_buffer*)atom; + + assert(cs->cdw + cb->atom.num_dw <= RADEON_MAX_CMDBUF_DWORDS); + memcpy(cs->buf + cs->cdw, cb->buf, 4 * cb->atom.num_dw); + cs->cdw += cb->atom.num_dw; +} + +void r600_init_command_buffer(struct r600_command_buffer *cb, unsigned num_dw, enum r600_atom_flags flags) +{ + cb->atom.emit = r600_emit_command_buffer; + cb->atom.num_dw = 0; + cb->atom.flags = flags; + cb->buf = CALLOC(1, 4 * num_dw); + cb->max_num_dw = num_dw; +} + +void r600_release_command_buffer(struct r600_command_buffer *cb) +{ + FREE(cb->buf); +} + static void r600_emit_surface_sync(struct r600_context *rctx, struct r600_atom *atom) { struct radeon_winsys_cs *cs = rctx->cs; diff --git a/src/gallium/drivers/r600/r600d.h b/src/gallium/drivers/r600/r600d.h index 3c3238a1242..40a0db0cf7c 100644 --- a/src/gallium/drivers/r600/r600d.h +++ b/src/gallium/drivers/r600/r600d.h @@ -36,9 +36,7 @@ #define EG_BOOL_CONST_OFFSET 0x0003A500 #define EG_BOOL_CONST_END 0x0003A506 -#define R600_CONFIG_REG_OFFSET 0X00008000 #define R600_CONFIG_REG_END 0X0000AC00 -#define R600_CONTEXT_REG_OFFSET 0X00028000 #define R600_CONTEXT_REG_END 0X00029000 #define R600_ALU_CONST_OFFSET 0X00030000 #define R600_ALU_CONST_END 0X00032000 @@ -158,7 +156,6 @@ #define PKT3_IT_OPCODE_C 0xFFFF00FF #define PKT3_PRED_S(x) (((x) >> 0) & 0x1) #define PKT0(index, count) (PKT_TYPE_S(0) | PKT0_BASE_INDEX_S(index) | PKT_COUNT_S(count)) -#define PKT3(op, count, predicate) (PKT_TYPE_S(3) | PKT3_IT_OPCODE_S(op) | PKT_COUNT_S(count) | PKT3_PRED_S(predicate)) /* Registers */ #define R_008490_CP_STRMOUT_CNTL 0x008490 -- 2.30.2